Evidence for a Crucial Role Played by Oxygen...
Transcript of Evidence for a Crucial Role Played by Oxygen...
Memory Devices
Evidence for a Crucial Role Played by Oxygen Vacancies in LaMnO 3 Resistive Switching Memories
Zhong-tang Xu , Kui-juan Jin , * Lin Gu , Yu-ling Jin , Chen Ge , Can Wang , Hai-zhong Guo , Hui-bin Lu , Rui-qiang Zhao , and Guo-zhen Yang
LaMnO 3 (LMO) fi lms are deposited on SrTiO 3 :Nb (0.8 wt%) substrates under various oxygen pressures to obtain different concentrations of oxygen vacancies in the fi lms. The results of X-ray diffraction verify that with a decrease of the oxygen pressure, the c -axis lattice constant of the LMO fi lms becomes larger, owing to an increase of the oxygen vacancies. Aberration-corrected annular-bright-fi eld scanning transmission electron microscopy with atomic resolution and sensitivity for light elements is used, which clearly shows that the number of oxygen vacancies increases with the decrease of oxygen pressure during fabrication. Correspondingly, the resistive switching property becomes more pronounced with more oxygen vacancies in the LMO fi lms. Furthermore, a numerical model based on the modifi cation of the interface property induced by the migration of oxygen vacancies in these structures is proposed to elucidate the underlying physical origins. The calculated results are in good agreement with the experimental data, which reveal from a theoretical point of view that the migration of oxygen vacancies and the variation of the Schottky barrier at the interface with applied bias dominate the resistive switching characteristic. It is promising that the resistive switching property in perovskite oxides can be manipulated by controlling the oxygen vacancies during fabrication or later annealing in an oxygen atmosphere.
1. Introduction
Conventional semiconductor memory technologies, including
dynamic random-access memory (DRAM) and fl ash memo-
ries, are rapidly approaching their bottleneck to continue
scaling down beyond a 16 nm threshold, [ 1 ] due to the fact that
the operation of such devices is established on charge storage
which leads to major diffi culties in information storage and
© 2012 Wiley-VCH Verlag GmbHsmall 2012, 8, No. 8, 1279–1284
DOI: 10.1002/smll.201101796
Dr. Z. Xu , Prof. K. Jin , Prof. L. Gu , Dr. Y. Jin , Dr. C. Ge , Dr. C. Wang , Dr. H. Guo , Prof. H. Lu , Dr. R. Zhao , G. Yang Beijing National Laboratory for Condensed Matter Physics Institute of Physics Chinese Academy of Science Beijing 100190, China E-mail: [email protected]
retrieval for the shrinking cells. [ 2 ] Motivated by the explora-
tion of alternatives with improved functionality over conven-
tional memories, extensive investigations of intriguing novel
memory concepts have been carried out up to now, which
range from phase-change RAM (PCRAM) [ 3 , 4 ] and ferroelec-
tric-related RAM [ 5–8 ] (FeRAM) to resistive switching RAM
(RRAM). [ 9–16 ] Among all these proposed concepts, robust
RRAM with great potential for high-density, high-speed,
nondestructive readout and nonvolatility is one of the most
promising candidates for the next-generation memories and
has attracted considerable attention. [ 17 ]
In general, the RRAM device has a simple capacitor-
like structure with a metal/insulator (or semiconductor)/
metal (MIM) stack. The basic operating principle is that
the resistance state can be switched between high and low
states with the application of an external electric stimulus.
A large variety of materials have shown resistive switching
properties. [ 17 ] In particular, transition metal oxides, [ 18 , 19 ] in
1279 & Co. KGaA, Weinheim wileyonlinelibrary.com
Z. Xu et al.
128
full papers
Figure 1 . SXRD patterns for LMO thin fi lms fabricated under oxygen pressures of a–c) 10, 5 × 10 − 2 , and 5 × 10 − 4 Pa, respectively.
which complex combinations of charge-, spin-, and orbital-
ordered states appear, [ 20 ] occupy an important position in this
emerging fi eld, [ 21 ] offering opportunities for further integra-
tion of potential multifunctionality into one device.
At present, a common perspective for the underlying
physical origins of the resistive switching phenomenon in
transition metal oxides is that the migration of oxygen vacan-
cies (or oxygen ions) plays a dominant role, thereby giving rise
to the modifi cation of the interface barrier [ 18 ] or inducing the
occurrence of the redox process. [ 22 ] So far, the important role
played by oxygen vacancies in resistive switching has been
widely discussed in many previous works. [ 23–32 ] However, it
would be more attractive and convincing to directly map out
the oxygen vacancies by an electron microscope technique
with a high resolution, and more direct proof of whether
oxygen vacancies are responsible for resistive switching
would be crucial for further development of RRAM. [ 17 ] Thus,
much effort is still needed to advance toward real-product
implementation, especially after clearly clarifying the funda-
mental origins.
Herein, we present a comparative study carried out
systematically to reveal the crucial role played by oxygen
vacancies in the resistive switching property in LaMnO 3
(LMO) fi lms. LMO fi lms, a typical perovskite oxide material,
were deposited on SrTiO 3 :Nb (SNTO, 0.8 wt%) substrates
under various oxygen pressures to obtain LMO fi lms with
different concentrations of oxygen vacancies. The results
from synchrotron-based X-ray diffraction (SXRD) veri-
fi ed that with a decrease of the oxygen pressure the c -axis
lattice constant of LMO fi lms becomes larger, owing to an
increase of the oxygen vacancies. To further characterize
the concentration of oxygen vacancies and the structural
evolution of these LMO fi lms at atomic sensitivity, sophis-
ticated aberration-corrected scanning transmission electron
microscopy (STEM) was utilized, which directly confi rmed
that with a decrease of the oxygen pressure during fabrica-
tion, the number of oxygen vacancies increased. Meanwhile,
measurements on the resistive switching showed that a
larger variation of resistance can be obtained with a forward
and reverse bias in LMO fi lms with more oxygen vacancies.
It is thus concluded that oxygen vacancies play a dominant
role in RRAM devices composed of transition metal oxides.
Furthermore, a numerical model based on modifi cation of
the interface property induced by the migration of oxygen
vacancies is proposed to elucidate the underlying physical
origins of resistive switching. Based on the present results,
it is promising that the RRAM property in the perovskite
oxides could be manipulated by controlling the oxygen
vacancies during fabrication or later annealing in an oxygen
atmosphere.
2. Results and Discussion
Figure 1 shows the results of SXRD of LMO fi lms fabricated
under various oxygen pressures. The diffraction peaks of
the LMO fi lms shift towards smaller diffraction angles with
a decrease of the oxygen pressure. The calculated c -axis lat-
tice constants of the LMO fi lms fabricated under oxygen
0 www.small-journal.com © 2012 Wiley-VCH V
pressures of 10, 5 × 10 − 2 , and 5 × 10 − 4 Pa are 3.892, 3.973, and
4.014 Å, respectively. We attribute the modifi cation of the
lattice constants to the infl uence of the oxygen vacancies,
naturally produced during the laser molecular beam epitaxy
(LMBE) processes. [ 33 ] Manganese ions vary from high to low
valence states due to the existence of oxygen vacancies, thus
leading to a larger manganese ionic radius and a larger lattice
constant. [ 34 ] It is worth mentioning that similar phenomena of
enlarged lattice constants induced by oxygen vacancies have
also been observed in previous works. [ 34–36 ] In other words, a
larger c -axis lattice constant corresponds to a larger number
of oxygen vacancies, which could be induced by structural
distortions in the thin fi lms fabricated under a lower oxygen
pressure.
Recent developments of aberration-corrected electron
optics have signifi cantly enhanced microscope performance,
thus enabling identifi cation of individual light atoms such as
oxygen. [ 37–39 ] In this study, we exploited an aberration-cor-
rected annular-bright-fi eld (ABF) imaging technique, which
is extremely sensitive to the presence of oxygen vacancies.
Figure 2 a–c display typical cross-sectional ABF micrographs
obtained from the LMO/SNTO junction viewed along the
[001] axis. La as well as Mn and O sites are clearly visible,
which can be distinguished according to the image contrast.
La is represented as the darkest spots, whereas O is the
lightest. These images reveal that the LMO fi lm was epi-
taxially grown on the c -SNTO substrate. Under extreme O
depletion conditions, however, the LMO thin fi lm exhibits
substantial structural distortions, possibly leading to the
formation of lattice defects as shown in Figure 2 c. Line
profi les along the vertical direction with the markers exhib-
ited in Figure 2 a–c are shown in Figure 2 d–f, respectively.
Note that the O sites are marked with red arrows and the
red/blue disks display stacking sequences of O–Mn with
erlag GmbH & Co. KGaA, Weinheim small 2012, 8, No. 8, 1279–1284
Role of Oxygen Vacancies in LaMnO3 Resistive Switching Memories
Figure 3 . Typical I–V curves for Pt/LMO/SNTO devices under oxygen pressures of 10, 5 × 10 − 1 , 5 × 10 − 2 , 5 × 10 − 3 , and 5 × 10 − 4 Pa. The direction of bias sweeping is indicated by arrows. The inset shows the measurement confi guration of the devices.
Figure 2 . a–c) ABF micrographs of the LMO/SNTO interfaces along the [001] axis corresponding to oxygen pressures of 10, 5 × 10 − 2 , and 5 × 10 − 4 Pa, respectively. Note that La is represented by the darkest spots, whereas O is the lightest. d–f) Line profi les along the vertical direction with respect to the markers shown in (a–c). The O sites are marked with red arrows. The red/blue disks display stacking sequences of O–Mn; lighter red ones stand for the presence of O vacancies. g–i) Line profi les along the horizontal direction with respect to the markers shown in (a–c).
lighter red ones standing for the presence of O vacancies.
Line profi les along the horizontal direction are shown in
Figure 2 g–i, together with the profi les in Figure 2 d–f. The
oxygen vacancy concentration is possibly controlled by
the growth kinetics. [ 40,41 ] Herein, it is confi rmed that the
number of oxygen vacancies increases with a decrease of
the oxygen pressure for our samples, consistent with the
SXRD measurements.
Figure 3 displays typical measured current–voltage ( I – V )
characteristics of LMO fi lms under various oxygen pres-
sures with Pt as top electrode (TE), and the inset exhibits
the schematic setup for I– V measurements. The forward bias
is defi ned by the current fl owing from the TE to the SNTO
bottom electrode (BE). The voltage bias was scanned as 1 →
2 → 3 → 4, as shown in Figure 3 . Obviously, the switching
hysteresis under forward bias is more pronounced than
that under reverse bias. A forward bias sweeping tunes the
memory into a high-resistance state (HRS or OFF), while
a reverse bias sweeping switches the memory into a low-
resistance state (LRS or ON). Figure 3 clearly shows that a
larger loop of I–V curve can be obtained in LMO fi lms fabri-
cated under a lower oxygen pressure. The resistance changes
© 2012 Wiley-VCH Verlag GmbHsmall 2012, 8, No. 8, 1279–1284
under the electric fi eld, which demonstrates that larger resis-
tive switching can be obtained in the LMO fi lms with more
oxygen vacancies under a sustained bias.
1281www.small-journal.com & Co. KGaA, Weinheim
Z. Xu et al.
12
full papers
Figure 4 . Pulse measurements of the Pt/LMO/SNTO devices with a fabrication oxygen pressure of 5 × 10 − 4 Pa. The pulse condition for the HRS is + 5 V with a pulse width of 1 ms and for the LRS is –5 V with a pulse width of 1 ms. The resistance was read out at 0.1 μ A.
To evaluate the stability of the devices, we performed
pulse endurance measurements. The results for the Pt/LMO/
SNTO structure fabricated under an oxygen pressure of
5 × 10 − 4 Pa are shown in Figure 4 . Bistable HRS/LRS
switching was achieved by applying a pulse voltage of + 5 and
–5 V with a 1 ms width for reset (LRS to HRS) and set (HRS
to LRS) operations. The resistance was then read out with a
read current of 0.1 μ A. It was observed that stable HRS and
LRS states were achieved with an R OFF / R ON ratio of about 18,
and no signifi cant change of the resistance ratio was observed
during pulse measurements. With further increase of the
number of cycles, the devices show no signifi cant resistance
decay in both the HRS and LRS.
From Figure 5 we can see that the R OFF / R ON ratio indi-
cated by numbers becomes larger with the decrease of the
oxygen pressure, except for the LMO fi lm with the largest
number of oxygen vacancies, that is, the fi lm fabricated under
5 × 10 − 4 Pa oxygen pressure. We think this is due to the exces-
sive structural distortion associated with too many oxygen
vacancies for the LMO fi lms fabricated under 5 × 10 − 4
82 www.small-journal.com © 2012 Wiley-VCH V
Figure 5 . Dependence of the HRS and LRS on the fabrication oxygen pressure. The R OFF / R ON ratios are indicated by numbers.
Pa oxygen pressure, which can be verifi ed by the ABF micro-
graph shown in Figure 2 c. According to the detailed char-
acterization of SXRD and aberration-corrected STEM, it
is known that for the structures fabricated under a lower
oxygen pressure there are more oxygen vacancies. Therefore,
more free carriers and mobile oxygen vacancies are obtained
for the fi lms fabricated under a lower oxygen pressure, which
induce a larger current and R OFF / R ON ratio. In other words,
the resistive switching property becomes more pronounced
in the fi lms containing more oxygen vacancies. This is direct
experimental evidence for the crucial role played by oxygen
vacancies in resistive switching.
To theoretically reveal the underlying physical origins, we
propose a numerical approach based on the oxygen vacancy
migration model. Detailed descriptions of the numerical
model can be found in the Supporting Information. The oxy-
gen-defi cient LMO is usually regarded as an n-type semicon-
ductor, because the oxygen vacancies act as donors in oxides
and become positively charged after releasing electrons
to the conduction band. [ 42 ] The work functions of Pt and
SNTO are about 5.3 [ 43 ] and 4.05 eV, [ 44 ] respectively. There-
fore, ohmic contact can be formed at the interface of LMO/
SNTO, and the Schottky barrier at the interface between Pt
and LMO will dominate the transport property of the entire
system. Hence, it is proposed that there exists an evolution
of several states for Pt/LMO/SNTO according to the number
of oxygen vacancies at the Pt/LMO Schottky interface under
the electric fi eld. When a negative voltage is applied on the
TE, the oxygen vacancies will accumulate at the Pt/LMO
interface, which makes the doping density at the interface
increase, while the oxygen vacancies will migrate away from
the Pt/LMO interface, thereby giving rise to a decrease of
doping density under the forward bias. The variation of the
carrier concentration due to the oxygen vacancy migration
will result in different states at the Pt/LMO Schottky barrier,
which causes high and low resistance states.
Because of the unobvious resistive switching property
under the reverse bias, only one state is taken into account in
the reverse bias for simplicity. Thus, we regard the system as
an evolution of three states, S1, S2, and S3, in the whole bias
sweeping process as shown in Figure 6 . The necessary param-
eters for calculations are listed in Table 1 . The doping density
of as-grown LMO (5 × 10 − 3 Pa) fi lm is measured as 3.5 × 10 19
cm − 3 with a Quantum Design physical property measurement
system (PPMS-9). The bandgap of the LMO fi lm is 1.6–2.0 eV
as reported by previous works. [ 45–48 ] With the decrease of
the doping density, the bandgap becomes larger according
to the well-known bandgap narrowing effect, [ 49 ] and the car-
rier mobility will become larger. The theoretical results are
in good agreement with the experimental data, as shown in
Figure 6 , which reveals from a theoretical point of view that
the migration of oxygen vacancies and the variation of the
Schottky barrier at the interface with applied bias dominate
the resistive switching property.
The band structures of S2 and S3, which are the states
of Pt/LMO/SNTO before and after applying forward bias,
respectively, are shown in Figure 7 . It can be seen that for
state S3 the width of the depletion region becomes wider and
the barrier height becomes larger compared with S2. In this
erlag GmbH & Co. KGaA, Weinheim small 2012, 8, No. 8, 1279–1284
Role of Oxygen Vacancies in LaMnO3 Resistive Switching Memories
Figure 7 . Calculated band structures of S2 and S3. E c , E v , and E f represent the conduction band, the valence band, and the Fermi level of LMO, respectively.
Figure 6 . Calculated I–V curves of states S1, S2, and S3 (solid lines) compared with measured I–V curves (open squares) of the structure fabricated under an oxygen pressure of 5 × 10 − 3 Pa. Arrows indicate the direction of bias sweeping.
simplifi ed model, we only consider the homogeneous oxygen
vacancy distribution at the Pt/LMO interface. The dynamic
inhomogeneous distributions of oxygen vacancies should be
further investigated for the resistive switching phenomenon
in the future.
3. Conclusion
We have designed a series of comparative experiments for
LMO fi lms fabricated under various oxygen pressures, and
proved a dominant role of oxygen vacancies in the resistive
switching property. The results of SXRD verifi ed that with
a decrease of the oxygen pressure the c -axis lattice con-
stant of LMO fi lms becomes larger, owing to an increase of
the oxygen vacancies. Aberration-corrected STEM was uti-
lized to further characterize the structure of the LMO fi lms
with atomic sensitivity, which directly confi rmed that with
a decrease of the oxygen pressure the number of oxygen
vacancies becomes larger. Correspondingly, larger resistive
switching can be obtained in LMO fi lms with more oxygen
vacancies under a sustained bias, while a consistent condition
is obtained under pulse bias except for the LMO fi lm with
the largest number of oxygen vacancies, the fi lm fabricated
under 5 × 10 − 4 Pa oxygen pressure, due to excessive structure
distortion associated with too many oxygen vacancies. The
resistive switching property becomes more pronounced with
LMO fi lms containing more oxygen vacancies. Furthermore,
© 2012 Wiley-VCH Verlag Gmbsmall 2012, 8, No. 8, 1279–1284
Table 1. Overview of the parameters used in the calculations.
S1 S2 S3
concentration [cm − 3 ] 9.8 × 10 19 3.5 × 10 19 5 × 10 18
mobility [cm 2 (V s) − 1 ] 1.5 2 7
bandgap [eV] 1.55 1.6 2.0
a numerical model based on the modifi cation of the interface
barrier induced by the migration of oxygen vacancies is pro-
posed to elucidate the physical origins. The calculated results
are in good agreement with the experimental data, which
reveal from a theoretical point of view that the migration of
oxygen vacancies and the variation of Schottky barrier at the
interface with applied bias dominate the resistive switching
characteristic. It is promising that the RRAM property in
perovskite oxides could be manipulated by controlling the
oxygen vacancies during fabrication or later annealing in an
oxygen atmosphere.
4. Experimental Section
Sample Preparation : LMO thin fi lms were epitaxially grown on SNTO (001) substrates with 3 × 5 mm size and 0.5 mm thickness by computer-controlled LMBE with a pulsed XeCl excimer laser beam ( ≈ 20 ns, 2 Hz, ≈ 1.5 J cm − 2 ) focused on a sintered ceramic LMO target. The SNTO substrates were carefully cleaned with alcohol, acetone, and deionized water before fabrication. Subse-quently, LMO fi lms with a thickness of 100 nm were deposited at 750 ° C under oxygen pressures of 5 × 10 − 4 , 5 × 10 − 3 , 5 × 10 − 2 , 5 × 10 − 1 and 10 Pa, and then annealed in situ under the same conditions for 20 min before cooling to room temperature in an oxygen atmosphere.
Characterization : The crystal structure was identifi ed by high-resolution synchrotron X-ray diffractometry by the BL14B1 beam line of Shanghai Synchrotron Radiation Facility (SSRF), using 1.24 Å X-rays with a Huber 5021 six-axes diffractometer.
Aberration-corrected STEM was performed using a 2100F trans-mission electron microscope operated at an acceleration voltage of 200 keV (JEOL, Tokyo, Japan). A CEOS probe-forming aberration corrector provided the ABF images with a spatial resolution better than 0.1 nm.
Electrical Characteristics : The dc and pulsed current–voltage ( I–V ) characteristics were measured by a Keithley 2400 source meter and Tektronix AFG320 arbitrary function generator at room temperature. The inset of Figure 3 shows the schematic setup for
1283www.small-journal.comH & Co. KGaA, Weinheim
Z. Xu et al.
1284
full papers
[ 1 ] International Technology Roadmap for Semiconductors, available at www.itrs.net.
[ 2 ] P. Zubko , J. M. Triscone , Nature 2009 , 460 , 45 . [ 3 ] M. H. R. Lankhorst , B.W. S. M. M. Ketelaars , R. A. M. Wolters ,
Nat. Mater. 2005 , 4 , 347 . [ 4 ] S. Lai , T. Lowrey , IEEE IEDM 2001 Tech. Dig. 2011 , 36 , 5 . [ 5 ] R. Meyer , R. Waser , J. Appl. Phys. 2006 , 100 , 051611 . [ 6 ] a) A. Q. Jiang , C. Wang , K. J. Jin , X. B. Liu , J. F. Scott , C. S. Hwang ,
T. A. Tang , H. B. Lu , G. Z. Yang , Adv. Mater. 2011 , 23 , 1277 ; b) C. Wang , K. J. Jin , Z. T. Xu , L. Wang , C. Ge , H. B. Lu , H. Z. Guo , M. He , G. Z. Yang , Appl. Phys. Lett. 2011 , 98 , 192901 .
[ 7 ] R. C. G. Naber , K. Asadi , P. W. M. Blom , D. M. de Leeuw , B. de Boer , Adv. Mater. 2010 , 22 , 933 .
[ 8 ] V. Garcia , S. Fusil , K. Bouzehouane , S. Enouz-Vedrenne , N. D. Mathur , A. Barthelemy , M. Bibes , Nature 2009 , 460 , 81 .
[ 9 ] A. Asamitsu , Y. Tomioka , H. Kuwahara , Y. Tokura , Nature 1997 , 388 , 50 .
[ 10 ] S. Q. Liu , N. J. Wu , A. Ignatiev , Appl. Phys. Lett. 2000 , 76 , 2749 . [ 11 ] K. Szot , W. Speier , G. Bihlmayer , R. Waser , Nat. Mater. 2006 , 5 ,
312 . [ 12 ] R. Waser , M. Aono , Nat. Mater. 2007 , 6 , 833 . [ 13 ] D. B. Strukov , G. S. Snider , D. R. Stewart , R. S. Williams , Nature
2008 , 453 , 80 . [ 14 ] Z. T. Xu , K. J. Jin , C. Wang , H. B. Lu , C. Wang , L. Wang , G. Z. Yang ,
Appl. Phys. A 2011 , 105 , 149 . [ 15 ] R. Muenstermann , T. Menke , R. Dittmann , R. Waser , Adv. Mater.
2010 , 22 , 4819 . [ 16 ] D. Strukov , J. Borghetti , R. Williams , Small 2009 , 5 , 1058 . [ 17 ] G. I. Meijer , Science 2008 , 319 , 1625 . [ 18 ] A. Sawa , Mater. Today 2008 , 11 , 28 . [ 19 ] S. F. Karg , G. I. Meijer , J. G. Bednorz , C. T. Rettner , A. G. Schrott ,
E. A. Joseph , C. H. Lam , M. Janousch , U. Staub , F. La Mattina , S. F. Alvarado , D. Widmer , R. Stutz , U. Drechsler , D. Caimi , IBM J. Res. Dev. 2008 , 52 , 481 .
[ 20 ] Y. Tokura , N. Nagaosa , Science 2000 , 288 , 462 . [ 21 ] H. Takagi , H. Y. Hwang , Science 2010 , 327 , 1601 .
I–V measurements. During the measurements, a bias was applied to the TE (Pt), whereas the BE (SNTO) was grounded. To protect devices, the applied voltage was restricted from –1.5 to + 3 V.
Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.
Acknowledgements
This work was supported by the National Basic Research Program of China (No. 2012CB921403) and the National Natural Science Foundation of China (No. 10825418 and No. 11134012). The XRD measurement was supported by the Shanghai Synchrotron Radia-tion Facility (SSRF).
www.small-journal.com © 2012 Wiley-VCH Ve
[ 22 ] R. Waser , R. Dittmann , G. Staikov , K. Szot , Adv. Mater. 2009 , 21 , 2632 .
[ 23 ] M. Janousch , G. I. Meijer , U. Staub , B. Delley , S. F. Karg , B. P. Andreasson , Adv. Mater. 2007 , 19 , 2232 .
[ 24 ] Y. B. Nian , J. Strozier , N. J. Wu , X. Chen , A. Ignatiev , Phys. Rev. Lett. 2004 , 92 , 178302 .
[ 25 ] K. Shibuya , R. Dittmann , S. B. Mi , R. Waser , Adv. Mater. 2010 , 22 , 411 .
[ 26 ] Y. B. Nian , J. Strozier , N. J. Wu , X. Chen , A. Ignatiev , Phys. Rev. Lett. 2007 , 98 , 146403 .
[ 27 ] C. Y. Lin , C. Y. Wu , C. Y. Wu , T. C. Lee , F. L. Yang , C. Hu , T. Y. Tseng , IEEE Electron Device Lett. 2007 , 28 , 366 .
[ 28 ] J. J. Yang , M. D. Pickett , X. M. Li , D. A. A. Ohlberg , D. R. Stewart , R. S. Williams , Nat. Nanotechnol. 2008 , 3 , 429 .
[ 29 ] S. Asanuma , H. Akohj , Phys. Rev. B 2009 , 80 , 235113 . [ 30 ] M. K. Yang , J. W. Park , T. K. Ko , J. K. Lee , Appl. Phys. Lett. 2009 ,
95 , 042105 . [ 31 ] D. S. Jeong , H. Schroeder , U. Breuer , R. Waser , J. Appl. Phys.
2008 , 104 , 123716 . [ 32 ] J. W. Parka , J. W. Park , K. Jung , M. K. Yang , J. K. Leeb , J. Vac. Sci.
Technol. B 2006 , 24 , 2205 . [ 33 ] W. Siemons , G. Koster , H. Yamamoto , W. A. Harrison , G. Lucovsky ,
T. H. Geballe , D. H. A. Blank , M. R. Beasley , Phys. Rev. Lett. 2007 , 98 , 196802 .
[ 34 ] L. Qiao , X. F. Bi , Thin Solid Films 2010 , 529 , 943 . [ 35 ] H. L. Cai , X. S. Wu , J. Gao , Chem. Phys. Lett. 2009 , 467 , 313 . [ 36 ] S. B. Mi , C. L. Jia , T. Heeg , O. Trithaveesak , J. Schubert , K. Urban ,
J. Cryst. Growth 2005 , 283 , 425 . [ 37 ] C. L. Jia , K. Urban , Science 2004 , 303 , 2001 . [ 38 ] S. D. Findlay , N. Shibata , H. Sawada , E. Okunishi , Y. Kondo ,
Y. Ikuhara , Ultramicroscopy 2010 , 110 , 903 . [ 39 ] O. L. Krivanek , M. F. Chisholm , V. Nicolosi , T. J. Pennycook ,
G. J. Corbin , N. Dellby , M. F. Murfi tt , Ch. S. Own , Z. S. Szilagyi , M. P. Oxley , S. T. Pantelides , S. J. Pennycook , Nature 2010 , 464 , 571 .
[ 40 ] S. Miyoshia , J. O. Honga , K. Yashir , A. Kaimaia , Y. Nigaraa , K. Kawamurab , T. Kawadaa , J. Mizusakia , Solid State Ionics 2003 , 161 , 209 .
[ 41 ] J. H. Kuo , H. U. Anderson , D. M. Sparlin , J. Solid State Chem. 1989 , 83 , 52 .
[ 42 ] S. R. Basu , L. W. Martin , Y. H. Chu , M. Gajek , R. Ramesh , R. C. Rai , X. Xu , J. L. Musfeldt , Appl. Phys. Lett. 2008 , 92 , 091905 .
[ 43 ] M. Minohara , I. Ohkubo , H. Kumigashira , M. Oshima , Appl. Phys. Lett. 2007 , 90 , 132123 .
[ 44 ] a) P. Han , K. J. Jin , H. B. Lu , Q. L. Zhou , Y. L. Zhou , G. Z. Yang , Appl. Phys. Lett. 2007 , 91 , 182102 ; b) K. J. Jin , H. B. Lu , Q. L. Zhou , K. Zhao , B. L. Cheng , Z. H. Chen , Y. L. Zhou , G. Z. Yang , Phys. Rev. B 2005 , 71 , 184428 ; c) C. L. Hu , K. J. Jin , P. Han , H. B. Lu , L. Liao , G. Z. Yang , Appl. Phys. Lett. 2008 , 93 , 162106 .
[ 45 ] T. Saitoh , A. E. Bocquet , T. Mizokawa , H. Namatame , M. Abbate , Y. Takeda , M. Takano , Phys. Rev. B 1995 , 51 , 13942 .
[ 46 ] D. Fuks , S. Dorfman , J. Felsteiner , L. Bakaleinikov , A. Gordon , E. A. Kotomin , Solid State Ionics 2004 , 173 , 107 .
[ 47 ] M. Hannani , D. Rached , M. Rabah , R. Khenata , N. Benayad , M. Hichour , A. Bouhemadou , Mater. Sci. Semicond. Process. 2008 , 11 , 81 .
[ 48 ] Y. Nohara , S. Yamamoto , T. Fujiwara , Phys. Rev. B 2009 , 79 , 195110 .
[ 49 ] S. M. Sze , Physica of Semiconductor Devices, Physics and Technology , 2nd ed., Wiley , New York 2001 .
Received: September 1, 2011 Published online: February 20, 2012
rlag GmbH & Co. KGaA, Weinheim small 2012, 8, No. 8, 1279–1284